Pranjal Singh

Pranjal Singh

Staff Data Scientist

About

Pranjal Singh is a Staff Data Scientist at udaan. He comes with a decade-long career in Data Science and AI, with a profound understanding across various AI & ML domains. His area of expertise extends to fraud prevention, AI-driven GTM design, recommendations, search algorithms, and route optimization. A holder of two patents and contributor to multiple academic publications in NLP and ML. Prior to joining udaan, India’s largest eB2B commerce platform, Pranjal held several key positions at Visa, Ola, and Ping Identity, providing him with a holistic perspective on the application of AI across different industry sectors and verticals. In the past, Pranjal has been a speaker at various industry forums. Pranjal has a Bachelor's and Master's degrees in Computer Science from IIT Kanpur.

Writing prompts manually is no longer enough for building robust AI applications at scale. This beginner-friendly (101) session introduces prompt and inference optimization frameworks designed to systematically improve LLM performance through programming abstractions instead of prompt hacking. We’ll explore how they help optimize prompts, reasoning chains, and model workflows automatically using metrics and feedback loops. The session covers core concepts like signatures, modules, optimizers, evaluation, and inference strategies, along with practical techniques to improve accuracy, consistency, latency, and cost. We’ll also discuss common production challenges such as hallucinations, brittle prompts, and scaling issues—ending with a live demo of building and optimizing an AI pipeline end-to-end.

Read More →